Light color acclimation is a key process in the globalocean distribution of Synechococcus cyanobacteriaThéophile Gréberta, Hugo Doréa, Frédéric Partenskya, Gregory K. Farranta,1, Emmanuel S. Bossb, Marc Picheralc,Lionel Guidic, Stéphane Pesantd,e, David J. Scanlanf, Patrick Winckerg, Silvia G. Acinash, David M. Kehoei,and Laurence Garczareka,2

aUMR 7144, Station Biologique, Sorbonne Université, CNRS, 29688 Roscoff Cedex, France; bMaine In-Situ Sound and Color Lab, University of Maine, Orono,ME 04469; cUMR 7093, Observatoire Océanologique, Sorbonne Université, CNRS, 06230 Villefranche-sur-Mer, France; dPANGAEA, Data Publisher for Earthand Environmental Science, University of Bremen, 28334 Bremen, Germany; eCenter for Marine Environmental Sciences, University of Bremen, 28359Bremen, Germany; fSchool of Life Sciences, University of Warwick, CV4 7AL Coventry, United Kingdom; gCommissariat à l’Energie Atomique et aux EnergiesAlternatives, Institut de Génomique, Genoscope, 91057 Evry, France; hDepartment of Marine Biology and Oceanography, Institute of Marine Sciences,Consejo Superior de Investigaciones Científicas, ES-08003 Barcelona, Spain; and iDepartment of Biology, Indiana University, Bloomington, IN 47405

Edited by David M. Karl, University of Hawaii, Honolulu, HI, and approved January 17, 2018 (received for review September 29, 2017)

Marine Synechococcus cyanobacteria are major contributors toglobal oceanic primary production and exhibit a unique diversityof photosynthetic pigments, allowing them to exploit a widerange of light niches. However, the relationship between pigmentcontent and niche partitioning has remained largely undetermineddue to the lack of a single-genetic marker resolving all pigmenttypes (PTs). Here, we developed and employed a robust methodbased on three distinct marker genes (cpcBA, mpeBA, and mpeW)to estimate the relative abundance of all known SynechococcusPTs from metagenomes. Analysis of the Tara Oceans datasetallowed us to reveal the global distribution of SynechococcusPTs and to define their environmental niches. Green-light special-ists (PT 3a) dominated in warm, green equatorial waters, whereasblue-light specialists (PT 3c) were particularly abundant in oligo-trophic areas. Type IV chromatic acclimaters (CA4-A/B), which areable to dynamically modify their light absorption properties tomaximally absorb green or blue light, were unexpectedly the mostabundant PT in our dataset and predominated at depth and highlatitudes. We also identified populations in which CA4 might benonfunctional due to the lack of specific CA4 genes, notably in warmhigh-nutrient low-chlorophyll areas. Major ecotypes within cladesI–IV and CRD1 were preferentially associated with a particular PT,while others exhibited a wide range of PTs. Altogether, this studyprovides important insights into the ecology of Synechococcus andhighlights the complex interactions between vertical phylogeny, pig-mentation, and environmental parameters that shape Synechococcuscommunity structure and evolution.

marine cyanobacteria | metagenomics | light quality | phycobilisome |Tara Oceans

Marine Synechococcus is the second most abundant phyto-plankton group in the world’s oceans and constitutes a

major contributor to global primary production and carbon cy-cling (1, 2). This genus displays a wide genetic diversity, andseveral studies have shown that, among the ∼20 clades definedbased on various genetic markers, five (clades I–IV and CRD1)predominate in situ and can be broadly associated with distinctsets of physicochemical parameters (3–5). In a recent study, wefurther defined ecologically significant taxonomic units (ESTUs),that is, organisms belonging to the same clade and co-occurringin the field, and highlighted that the three main parameters af-fecting the in situ distribution of these ESTUs were temperatureand availability of iron and phosphorus (6). However, marineSynechococcus also display a wide pigment diversity, suggestingthat light could also influence their ecological distribution, bothqualitatively and quantitatively (7, 8).This pigment diversity comes from differences in the composi-

tion of their main light-harvesting antennae, called phycobilisomes(PBSs) (7–9). These water-soluble macromolecular complexesconsist of a central core anchoring at least six radiating rods

made of several distinct phycobiliproteins, that is, proteins towhich specific enzymes (phycobilin lyases) covalently attachchromophores called phycobilins (7, 10). Although the PBS coreis conserved in all marine Synechococcus, rods have a very var-iable composition, and three main pigment types (PTs) areusually distinguished (Fig. S1) (7, 11). In PT 1, PBS rods aresolely made of phycocyanin (PC) (encoded by the cpcBA operon)and bear the red-light–absorbing phycocyanobilin (PCB) (Amax =620 nm) as the sole chromophore. In PT 2, rods are made of PCand phycoerythrin I (PE-I) (encoded by cpeBA) and attach bothPCB and the green-light–absorbing phycoerythrobilin (PEB)(Amax = 550 nm). All other marine Synechococcus belong to PT3 and have rods made of PC, PE-I, and PE-II (encoded bympeBA) that bind PCB, PEB, and the blue-light–absorbingphycourobilin (PUB) (Amax = 495 nm; Fig. S1). Several subtypescan be defined within PT 3, based on the fluorescence excitationratio at 495 and 545 nm (hereafter, Ex495:545; Fig. S1), a proxy forthe PUB:PEB ratio. This ratio is low (Ex495:545 < 0.6) in sub-type 3a (green-light specialists), intermediate in subtype 3b (0.6

Significance

Understanding the functional diversity of specific microbialgroups at the global scale is critical yet poorly developed. Bycombining the considerable knowledge accumulated throughrecent years on the molecular bases of photosynthetic pig-ment diversity in marine Synechococcus, a major phyto-planktonic organism, with the wealth of metagenomic dataprovided by the Tara Oceans expedition, we have been able toreliably quantify all known pigment types along its transectand provide a global distribution map. Unexpectedly, cellsable to dynamically change their pigment content to matchthe ambient light color were ubiquitous and predominated inmany environments. Altogether, our results unveiled the roleof adaptation to light quality on niche partitioning in a keyprimary producer.

Author contributions: T.G., F.P., and L. Garczarek designed research; T.G., H.D., F.P., E.S.B.,and L. Garczarek performed research; T.G., H.D., G.K.F., E.S.B., M.P., L. Guidi, S.P., D.J.S.,P.W., S.G.A., and L. Garczarek contributed new reagents/analytic tools; T.G., H.D., F.P.,G.K.F., E.S.B., and L. Garczarek analyzed data; and T.G., F.P., E.S.B., D.J.S., S.G.A., D.M.K.,and L. Garczarek wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.

Data deposition: The sequencing data and accession numbers are provided in Dataset S1.1Present address: Food Safety, Environment, and Genetics, Matís Ltd., 113 Reykjavík,Iceland.

2To whom correspondence should be addressed. Email: laurence.garczarek@sb-roscoff.fr.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1717069115/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1717069115 PNAS Latest Articles | 1 of 10

ENVIRONMEN

TAL

SCIENCE

SPN

ASPL

US

≤ Ex495:545 < 1.6), and high (Ex495:545 ≥ 1.6) in subtype 3c (blue-light specialists) (7, 11). Additionally, strains of subtype 3d areable to change their PUB:PEB ratio depending on ambient lightcolor, a process called type IV chromatic acclimation (hereafterCA4), allowing them to maximally absorb blue or green light(11–14). Comparative genomic analyses showed that genes in-volved in the synthesis and regulation of PBS rods are gatheredinto a dedicated genomic region, the content and organization ofwhich correspond to the different PTs (7). Similarly, chromaticacclimation has been correlated with the presence of a smallspecific genomic island (CA4 genomic island) that exists in twodistinct configurations (CA4-A and -B) (11). Both contain tworegulators (fciA and fciB) and a phycobilin lyase (mpeZ in CA4-A ormpeW in CA4-B), thus defining two distinct CA4 genotypes:3dA and 3dB, respectively (11, 14, 15). Finally, some strainspossess a complete or partial CA4 genomic island but are notable to perform CA4, displaying a fixed Ex495:545 correspondingto 3a, 3b, or 3c phenotypes (11).As there is no correspondence between pigmentation and core

genome phylogeny (7, 16, 17), deciphering the relative abun-dance and niche partitioning of Synechococcus PTs in the envi-ronment requires specific approaches. In the past 30 y, studieshave been based either on (i) proxies of the PUB:PEB ratio asassessed by flow cytometry (18–20), fluorescence excitationspectra (21–27), and epifluorescence microscopy (28); or (ii)phylogenetic analyses of cpcBA or cpeBA (17, 29–34). Thesestudies showed that PT 1 is restricted to and dominates in lowsalinity surface waters and/or estuaries, which are characterizedby a high turbidity resulting in a red-wavelengths–dominatedlight field (18, 22, 31–38), whereas PT 2 is found in coastal shelfwaters or in the transition zones between brackish and oceanicenvironments with intermediate optical properties (18, 27, 34,36–39). Finally, PT 3 with increasing PUB:PEB ratio are foundover gradients from onshore mesotrophic waters, characterizedby green-light dominance, to offshore oligotrophic waters, whereblue light penetrates the deepest (19–24, 28, 36, 38, 40). Someauthors reported an increase in the PUB:PEB ratio with depth(19, 21, 24), while others observed a constant ratio throughoutthe water column, a variability potentially linked to the loca-tion, water column features, and/or environmental parameters(22, 25, 28).However, these analyses based on optical properties could

only describe the distribution of high- and low-PUB populationswithout being able to differentiate green- (3a) or blue-light (3c)specialists from CA4 cells (3d) acclimated to green or blue light,while genetic analysis solely based on cpcBA and/or cpeBA couldnot differentiate all PTs. For instance, only two studies havereported CA4 populations in situ either in the western EnglishChannel (17) or in subpolar waters of the western Pacific Ocean(29), but none of them was able to differentiate CA4-B fromhigh-PUB (i.e., 3c) populations. As a consequence, the globalrelative abundance of the different Synechococcus PTs, particu-larly CA4, and the link between genetic and pigment diversityhave remained largely unclear.Here, we analyzed 109 metagenomic samples collected from

all major oceanic basins during the 2.5-y Tara Oceans (2009–2011) expedition (41) using a bioinformatic pipeline combining ametagenomic read recruitment approach (6, 42) to recruit singlereads from multiple PBS gene markers and placement of thesereads in reference trees to assign them to a given PT. Thispipeline allowed a description of the worldwide distribution of allknown Synechococcus PTs, as well as of their realized environ-mental niches (sensu ref. 43). This study provides a synoptic viewof how a major photosynthetic organism adapts to natural lightcolor gradients in the ocean.

ResultsA Robust Approach for Estimating Pigment Type Abundance fromMetagenomes. We developed a multimarker approach com-bining phylogenetic information retrieved from three differ-ent genes or operons (cpcBA, mpeBA, and mpeW; Fig. 1 andDatasets S1 and S2) to overcome the issue of fully resolving thewhole range of PTs. While cpcBA discriminated PT 1, 2, and 3(Fig. 1A), only the mpeBA operon, a PT 3-specific marker, wasable to distinguish the different PT 3 subtypes (Fig. 1B), althoughas for cpeBA it could not differentiate PT 3dB (CA4-B) from PT3c (i.e., blue light specialists) (11, 29). The mpeW marker was thusselected to specifically target PT 3dB and, by subtraction, enu-merate PT 3c (Fig. 1C). Using the cpcBA marker, members of PT2 were split into two well-defined clusters, 2A and 2B (Fig. 1A),the latter corresponding to a purely environmental PT identifiedfrom assembled metagenomes of the Baltic Sea (38). StrainsKORDI-100 and CC9616 also clustered apart from other strainsin the mpeBA phylogeny, suggesting that they have a divergentevolutionary history from other PT 3 members (Fig. 1B). This issupported by the diverged gene content and order of their PBSrod genomic region, and these strains were recently referred to asPT 3f, even though they have a similar phenotype as PT 3c(Ex495:545 ratio ≥ 1.6) (30). To investigate the phylogenetic reso-lution of small fragments of these three markers, sequences wereremoved one at a time from the reference database, and simulatedreads (150 bp long compared with 164 bp in average for TaraOceans cleaned/merged reads) generated from this sequence wereassigned using our bioinformatic pipeline against a databasecomprising the remaining sequences. Inferred and known PTswere then compared. The percentage of simulated reads assignedto the correct PT was between 93.2% and 97.0% for all threemarkers, with less than 2.1–5.6% of reads that could not be clas-sified and an error rate below 2%, showing that all three markersdisplay a sufficient resolution to reliably assign the different PTs(Fig. S2 B, D, and F).To ensure that the different markers could be quantitatively

compared in a real dataset, we examined the correlations be-tween estimates of PT abundances using the different markers inthe 109 metagenomes analyzed in this study. Total cpcBA countswere highly correlated (R2 = 0.994, n = 109; Fig. S3A) with totalSynechococcus counts obtained with the petB gene, which waspreviously used to study the phylogeography of marine pico-cyanobacteria (6), and the correlation slope was not significantlydifferent from 1 (slope, 1.040; Wilcoxon’s paired difference testP value = 0.356). cpcBA is thus as good as petB at capturing thetotal population of Synechococcus reads. Moreover, counts ofcpcBA reads assigned to PT 3 and total mpeBA counts (specificfor PT 3) were also strongly correlated (R2 = 0.996, n = 109; Fig.S3B), and not skewed from 1 (slope of 0.991; Wilcoxon’sP value = 0.607). Although no redundant information for PT3dB is available with the three selected markers, another markertargeting 3dB (fciAB) was tested and produced results similar tompeW (Fig. S3C). These results demonstrate that our multi-marker approach can be used to reliably and quantitatively inferthe different Synechococcus PTs from short metagenomic reads,with PT 1, 2A, and 2B abundances being assessed by cpcBAnormalized counts; PT 3a, 3f, and 3dA by mpeBA normalizedcounts; PT 3dB by mpeW normalized counts; and PT 3c by thedifference between mpeBA normalized counts for 3c + 3dB andmpeW normalized counts. We thus used this approach on theTara Oceans metagenomes, generated from 109 samples col-lected at 65 stations located in the major oceanic basins (Fig. 2).

CA4 Populations Are Widespread and Predominate at Depth and HighLatitudes. The latitudinal distribution of Synechococcus inferredfrom cpcBA counts is globally consistent with previous studies (2,6, 44), with Synechococcus being present in most oceanic waters,

2 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1717069115 Grébert et al.

but quasi-absent (<20 cpcBA counts) beyond 60°S (SouthernOcean stations TARA_082 to TARA_085; Fig. 2B). Overall, thenumber of recruited cpcBA reads per station was between 0 and8,151 (n = 63; median, 449; mean, 924; SD, 1,478) for surfaceand 0 and 3,200 (n = 46; median, 170; mean, 446; SD, 664) fordeep chlorophyll maximum (DCM) samples, respectively. Sta-tions with less than 30 cpcBA reads were excluded fromfurther analysis.PT 1 and 2, both of which are known to be mostly found and

abundant in coastal waters (29, 36, 38, 45), were expectedly al-most absent from this dataset (total of 15 and 513 cpcBA reads,respectively; Fig. 2) since the Tara cruise sampling was princi-pally performed in oceanic waters. While PT 2A was mostlyfound at the surface at one station off Panama (TARA_141,

417 out of 6,637 reads at this station; Fig. 2B), PT 2B was vir-tually absent (total of 3 cpcBA reads) from our dataset and mightthus be confined to the Baltic Sea (38). This low abundance ofPT 1 and 2B precluded the correlation analysis between theirdistribution and physicochemical parameters. PT 3 was by far themost abundant along the Tara Oceans transect, accounting for99.1 ± 1.4% (mean ± SD) of cpcBA reads at stations with ≥30cpcBA read counts. Interestingly, several PT 3 subtypes often co-occurred at a given station.PT 3a (green-light specialists) totaled 20.3% of read counts,

with similar abundance between surface (20.5%) and DCM(19.4%) samples, and was particularly abundant in intertropicaloceanic borders and regional seas, including the Red Sea, theArabian Sea, and the Panama/Gulf of Mexico area (Fig. 2B).

A B

C

Fig. 1. Maximum-likelihood phylogenetic trees of (A) cpcBA operon, (B) mpeBA operon, and (C) the mpeW/Y/Z gene family. The cpcBA tree includes bothstrains with characterized pigment type (PT) and environmental sequences (prefixed with GS) assembled from metagenomes of the Baltic Sea (38). Circles atnodes indicate bootstrap support (black, >90%; white, >70%). Note that, for PT 2B clade, only environmental sequences are available. The PT associated witheach sequence is indicated as a colored square. The scale bars represent the number of substitutions per nucleotide position. Please note that there are two tofour copies of the cpcBA operon in PT 1, which are either identical or closely related.

Grébert et al. PNAS Latest Articles | 3 of 10

ENVIRONMEN

TAL

SCIENCE

SPN

ASPL

US

Correlation analyses show that this PT is consistently associatedwith high temperatures but also with greenish (as estimated froma low blue to green downwelling irradiance ratio, Irr495:545),particle-rich waters (high particle backscattering at 470 nm andbeam attenuation coefficient at 660 nm; Fig. 3). Still, in contrastwith previous studies that reported the distribution of low-PUBpopulations (19, 21, 23, 24, 26, 27), this PT does not seem to berestricted to coastal waters, explaining its absence of correlationwith chlorophyll concentration and colored dissolved organicmatter (cDOM).Blue light specialists (PT 3c) appear to be globally widespread,

with the exception of high-latitude North Atlantic waters, andaccounted for 33.4% of reads, with a higher relative abundanceat the surface (36.8%) than at the DCM (23.3%; Fig. 2A). ThisPT is dominant in transparent, oligotrophic, iron-replete areassuch as the Mediterranean Sea as well as South Atlantic andIndian Ocean gyres (Figs. 2B and 4C). In the South Pacific, PT3c was also found to be predominant in the Marquesas Islandsarea (TARA_123 and 124), where the coast proximity induced alocal iron enrichment (6). Consistently, PT 3c was found to bepositively associated with iron concentration, high temperature,and DCM depth, and anticorrelated with chlorophyll fluores-cence, nitrogen concentrations, net primary production (NPP),as well as other related optical parameters, such as backscat-tering at 470 nm and beam attenuation coefficient at 660 nm(Fig. 3). Despite its rarity, PT 3f seems to thrive in a similarenvironment, with the highest relative abundances in the IndianOcean and Mediterranean Sea (Figs. 2B and 4C). Its occurrence

in the latter area might explain its strong anticorrelation withphosphorus availability.Both CA4 types, 3dA and 3dB, which represented 22.6% and

18.9% of reads, respectively, were unexpectedly widespread andcould locally account for up to 95% of the total Synechococcuspopulation (Figs. 2 and 4C and Fig. S4). In contrast to blue- andgreen-light specialists, both CA4 types were proportionally lessabundant at the surface (19.8% and 17.5%, for 3dA and 3dB,respectively) than at depth (30.9% and 22.9%). Interestingly, PT3dA and 3dB generally displayed complementary distributionsalong the Tara Oceans transect (Fig. 2B). PT 3dA was pre-dominant at high latitude in the Northern Hemisphere as well asin other vertically mixed waters such as in the Chilean upwelling(TARA_093) or in the Agulhas current (TARA_066 and 68; Fig.2B). Accordingly, PT 3dA distribution seems to be driven by low-temperature, high-nutrient, and highly productive waters (highNPP, chlorophyll a, and optical parameters), a combination ofphysicochemical parameters almost opposite to those observedfor blue-light specialists (PT 3c; Fig. 3). In contrast, PT 3dBshares a number of characteristics with PT 3c, including theanticorrelation with nitrogen concentration and association withiron availability (as indicated by both a positive correlation with[Fe] and negative correlation with the iron limitation proxy Φsat;Fig. 3), consistent with their widespread occurrence in iron-replete oceanic areas. Also noteworthy, PT 3dB was one of thesole PTs (with 3f) to be associated with low photosyntheticallyavailable radiation (PAR).

0 10 20 30 40 50 60 70 80 90 100

A

B

Fig. 2. Distribution of Synechococcus pigment types (PTs). (A) Relative abundance of each PT in the whole dataset (Total), in surface, and at the deepchlorophyll maximum (DCM). (B) Map showing the global distribution of all Synechococcus PTs in surface waters along the Tara Oceans transect. Diameters ofpies are proportional to the number of cpcBA reads normalized by the sequencing effort. Stations with less than 30 cpcBA or mpeBA reads are indicated byopen circles, and those with no cpcBA reads by black dots. Numbers next to pies correspond to Tara Oceans stations.

4 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1717069115 Grébert et al.

Niche Partitioning of Synechococcus Populations Relies on a SubtleCombination of ESTU and PT Niches. We previously showed thattemperature and iron and phosphorus availability constitutedmajor factors influencing the diversification and niche partitioningof Synechococcus ESTUs (i.e., genetically related subgroups withinclades that co-occur in the field) (6). However, these resultscannot be extended to PTs since the pigment content does notfollow the vertical phylogeny (7). To decipher the respective rolesof genetic and pigment diversity in Synechococcus communitystructure, we examined the relationships between ESTUs and PTsin situ abundances through correlation and nonmetric multidi-mensional scaling (NMDS) analyses (Fig. 4 A and B) and com-pared their respective distributions (Fig. 4C and Fig. S4).Interestingly, all PTs are either preferentially associated with

or excluded from a subset of ESTUs. PT 2A is found at lowabundance at a few stations along the Tara Oceans transect, and,when present, it is seemingly associated with the rare ESTU 5.3B(Fig. 4A), an unusual PT/ESTU combination so far only ob-served in metagenomes from freshwater reservoirs (46). PT 3a isassociated with ESTUs EnvBC (occurring in low-iron areas) andIIA, the major ESTU in the global ocean (Fig. 4A), a resultconsistent with NMDS analysis, which shows that PT 3a is foundin assemblages dominated by these two ESTUs (indicated by redand gray backgrounds in Fig. 4B), as well as with independentobservations on cultured strains (Dataset S3). PT 3c is associatedwith ESTU IIIA (the dominant ESTU in P-depleted areas), asobserved on many isolates (Dataset S3), and is also linked, likePT 3f, with ESTUs IIIB and WPC1A, both present at lowerabundance than IIIA in P-poor waters (Fig. 4A). PT 3f is alsoassociated with the newly described and low-abundance ESTUXXA (previously EnvC; Fig. S5) (4, 6). Both PT 3f and ESTUXXA were rare in our dataset but systematically co-occurred, inagreement with the fact that the only culture representative ofthe latter clade belongs to PT 3f (Dataset S3).PT 3dA appears to be associated with all ESTUs from clades

CRD1 (specific to iron-depleted areas) as well as with thoserepresentative of coastal and cold waters (IA, IVA, IVC) but isanticorrelated with most other major ESTUs (IIA, IIIA and -B,WPC1A, and 5.3B; Fig. 4A). This pattern is opposite to PT 3dB,which is preferentially found associated with ESTU IIA, IIB, and

5.3A, but not in CRD1A or -C (Fig. 4A). Thus, it seems that thetwo types of CA4 are found in distinct and complementary setsof ESTUs. Interestingly, our analysis might suggest the occur-rence of additional PTs not isolated so far, since a number ofreads (0.7% and 2.7% of cpcBA and mpeBA counts, respectively;Fig. 2A) could not be assigned to any known PTs. For instance,while most CRD1C seem preferentially associated with PT 3dA,a fraction of the population could only be assigned at the PT3 level (Fig. 4A). Similarly, a number of reads could not beassigned to any known PT in stations rich in ESTU 5.3A andXXA, although one cannot exclude that this observation mightbe due to a low number of representative strains, and thus PTreference sequences, for these ESTUs.The preferred association of PTs with specific ESTUs is also

well illustrated by some concomitant shifts of PTs and ESTUassemblages. For instance, in the wintertime North AtlanticOcean, the shift from 3dB-dominated stations on the westernside (TARA_142 and TARA_146 to 149) to 3dA-dominatedstations near European coasts (TARA_150 to 152) and North ofGulf Stream (TARA_145) is probably related to the shift inESTU assemblages occurring along this transect, with ESTU IIAbeing gradually replaced by ESTU IVA (Fig. 4C) (6). Simi-larly, the takeover of CRD1C by IIA in the Marquesas Islandarea (TARA_123 to 125), which is iron-enriched with regard tosurrounding high-nutrient low-chlorophyll (HNLC) waters(TARA_122 and 128), perfectly matched the corresponding re-placement of PT 3dA by 3c. However, in several other cases, PTshifts were not associated with a concomitant ESTU shift or viceversa. One of clearest examples of these dissociations is thetransect from the Mediterranean Sea to the Indian Ocean, wherethe entry in the northern Red Sea through the Suez Canal trig-gered a sharp shift from a IIIA- to a IIA-dominated community(TARA_030 and 031), which was not accompanied by any ob-vious change in PTs. Conversely, a sharp rise in the relativeabundance of PT 3a was observed in the southern Red Sea/northeastern Indian Ocean (TARA_033 to 038) without changesin the large dominance of ESTU IIA. Altogether, this stronglysuggests that a subtle combination of ESTUs and PTs respectiveniche occupancy is responsible for the observed niche parti-tioning of Synechococcus populations.

Fig. 3. Correlation analysis between Synechococcus pigment types (PTs) and environmental parameters measured along the Tara Oceans transect for allsampled depths. The scale shows the degree of correlation (blue) or anticorrelation (red) between two variables. Nonsignificant correlations (adjusted P value > 0.05)are indicated by crosses. Number of observations for each environmental parameter is indicated at Bottom. Abbreviations: BAC, beam attenuation coefficient;Backscatt., backscattering; cDOM fluo., colored dissolved organic matter fluorescence; DCM, deep chlorophyll maximum; Irr495:545, ratio of downwelling irradiance at495 and 545 nm; IS, in situ; MLD, mixed layer depth; NPP, net primary production; PAR, photosynthetically active radiation; part., particulate; Φsat, satellite-basednonphotochemical quenching (NPQ)-corrected quantum yield of fluorescence (proxy for iron limitation) (6).

Grébert et al. PNAS Latest Articles | 5 of 10

ENVIRONMEN

TAL

SCIENCE

SPN

ASPL

US

Deficient Chromatic Acclimaters Are Dominant in HNLC Areas. Al-though our results clearly indicate that CA4 cells represent alarge proportion of the Synechococcus community in a widerange of ecological niches, this must be somewhat tempered bythe fact that, in culture, about 30% of the strains possessing aCA4-A or -B genomic island are not able to chromatically ac-

climate (Dataset S3) (11). Some of these natural mutants havean incomplete CA4 genomic island (Fig. S6K). For example,strains WH8016 (ESTU IA) and KORDI-49 (WPC1A) both lackthe CA4-A–specific lyase-isomerase MpeZ, an enzyme shown tobind a PUB molecule on PE-II (14), and display a green-lightspecialist phenotype (PT 3a, Ex495:545 ∼ 0.4) whatever the

A B

C

Fig. 4. Relationship between Synechococcus pigment types (PTs) and ecologically significant taxonomic units (ESTUs) (as defined in ref. 6). (A) Correlation analysisbetween Synechococcus PTs and the most abundant ESTUs (>1% relative abundance) for all sampled depths (the complete dataset is shown in Fig. S5). Non-significant correlations (adjusted P value > 0.05) are indicated by crosses. The surface of station TARA_067, identified as an outlier (Fig. S7), was removed for thisanalysis. (B) NMDS analysis of stations according to Bray–Curtis distance between PT assemblages. Samples that belong to the same ESTU assemblage have beencontoured with a background color according to the color code used in ref. 6, namely, red, assemblage 1 dominated by ESTU IIA; yellow, assemblage 2 dominatedby ESTU IIIA; dark blue, assemblage 4 dominated by ESTUs IA and IVA/B; pink, assemblage 5 codominated by ESTUs IIB and IVA/B; gray, assemblage 6 codominatedby ESTUs CRD1C and EnvBC; light blue, assemblage 8 codominated by ESTUs IVA/B, EnvBB, and CRD1A/B. (C) PT and ESTU relative abundance at each surfacestation along the Tara Oceans transect. Oceanic provinces are indicated in the Top gray panels. IO, Indian Ocean; MS, Mediterranean Sea; NAO, North AtlanticOcean; NPO, North Pacific Ocean; RS, Red Sea; SAO, South Atlantic Ocean; SO, Southern Ocean; SPO, South Pacific Ocean.

6 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1717069115 Grébert et al.

ambient light color (11). However, since they possess a PT 3ampeBA allele, reads from field WH8016- or KORDI-49–likecells are adequately counted as PT 3a (Fig. S6K). Another CA4-deficient strain, BIOS-E4-1 (ESTU CRD1C), possesses mpeZand a 3dA mpeBA allele but lacks the CA4 regulators FciA andFciB as well as the putative lyase MpeY and exhibits a fixed blue-light specialist phenotype (PT 3c, Ex495:545 ∼ 1.7; Fig. S6K) (11,15). Thus, reads from such natural Synechococcus CA4-incapablemutants in the field are counted as 3dA using the mpeBAmarker. Last, the strain MVIR-18-1 possesses a complete CA4-Aisland and a 3dA mpeBA allele but lacks mpeU, a gene necessaryfor blue-light acclimation (Fig. S6K) (47). While MVIR-18-1 displays a fixed green-light phenotype, reads from such Syn-echococcus are also erroneously counted as 3dA.To assess the significance of these genotypes in the field, we

compared the normalized read counts obtained for 3dA withmpeBA, fciAB, mpeZ, mpeU, and mpeY (Fig. S6 A–J). Overall,this analysis revealed a high consistency between these differ-ent markers (0.860 < R2 < 0.986), indicating that most mpeZ-containing populations also contained 3dA alleles for fciAB,mpeY, mpeU, and mpeBA and are therefore likely able to per-form CA4. However, a number of stations, all located in HNLCareas (TARA_094, 111, and 122–128 in the Pacific Ocean andTARA_052 located northwest of Madagascar; Fig. 2B), dis-played more than 10-fold higher mpeBA, mpeU, and mpeZcounts than fciAB and mpeY counts (Fig. S6 A, B, E, F, H, and I).This indicates that a large proportion or even the whole population(TARA_122 and 124) of 3dA in these HNLC areas is probablylacking the FciA/B regulators and MpeY and, like strain BIOS-E4-1 (Fig. S6K), might thus be stuck in the blue-light specialist phe-notype (PT 3c) (11). Conversely, station TARA_067 exhibitedconsistently more than twice the fciAB andmpeZ counts comparedwith mpeBA, mpeY, or mpeU (Fig. S6 B–E, G, and H) and was aclear outlier when comparing pigment type and clade composition(Fig. S7). This suggests that the proportion of PT 3dA might havebeen underestimated at this station, as a significant proportion ofthis population probably corresponds to PT 3a genotypes that haveacquired a CA4-A island by lateral gene transfer, as is seeminglythe case for strains WH8016 and KORDI-49. Finally, no stationexhibited markedly lower mpeU counts compared with all othergenes, indicating that the genotype of strain MVIR-18-1 is prob-ably rare in the oceans.It must be noted that two out of the six sequenced CA4-B

strains (WH8103 and WH8109) also have a deficient CA4 phe-notype and display a constant, intermediate Ex495:545 ratio(0.7 and 1, respectively), despite any obvious PBS- or CA4-related gene deletion (11). Accordingly, the plot of 3dB nor-malized read counts obtained with mpeW vs. fciAB shows noclear outlier (Fig. S3C).

DiscussionMarine Synechococcus display a large pigment diversity, withdifferent PTs preferentially harvesting distinct regions of thelight spectrum. Previous studies based on optical properties oron a single genetic marker could not differentiate all PTs (17,29–31), and thus neither assess their respective realized envi-ronmental niches (43) nor the role of light quality on the relativeabundance of each PT. Here, we showed that a metagenomicread recruitment approach combining three genetic markers canbe used to reliably predict all major PTs. Applied to the exten-sive Tara Oceans dataset, this original approach, which avoidsPCR amplification and cloning biases, allowed us to describe thedistribution of the different Synechococcus PTs at the globalscale and to refine our understanding of their ecology.PT 3 was found to be largely dominant over PT 1 and 2 along

the oceanic Tara Oceans transect, in agreement with the coastal-restricted distribution of the latter PTs (18, 22, 27, 31–34, 37–39).Biogeography and correlation analyses with environmental pa-

rameters provided several important insights concerning nichepartitioning of PT 3 subtypes. Green- (PT 3a) and blue-light (PT3c) specialists were both shown to dominate in warm areas butdisplay clearly distinct niches, with 3a dominating in Synechococcus-rich stations located on oceanic borders, while 3c predominated inpurely oceanic areas where the global abundance of Synechococcusis low. These results are in agreement with the prevailing view of anincrease in the PUB:PEB ratio from green onshore mesotrophicwaters to blue offshore oligotrophic waters (19–24, 26–29, 40, 48).Similarly, we showed that PT 3dB, which could not be distin-guished from PT 3c in previous studies (17, 29–31), prevails inmore coastal and/or mixed temperate waters than do 3c pop-ulations. The realized environmental niche of the second type ofCA4 (PT 3dA) is the best defined of all PTs as it is clearly asso-ciated with nutrient-rich waters and with the coldest stations ofour dataset, occurring at high latitude, at depth, and/or in verti-cally mixed waters (e.g., TARA_068, 093, and 133). This result isconsistent with a recent study demonstrating the dominance of3dA in subarctic waters of the Northwest Pacific Ocean (29),suggesting that the prevalence of 3dA at high latitude can begeneralized. The decrease of PT 3c (blue-light specialists) withdepth is unexpected given previous reports of a constant (22, 25,28, 49) or increasing (19, 21, 24) PUB:PEB ratio throughout thewater column. However, the high abundance of CA4 can reconcilethese observations with the decreased abundance of PT 3c, as cellscapable of CA4 likely have a blue-light phenotype at depth. Al-together, while little was previously known about the abundanceand distribution of CA4 populations in the field, here we showthat they are ubiquitous, dominate in a wide range of niches, arepresent both in coastal and oceanic mixed waters, and overall arethe most abundant Synechococcus PT.The relationship between ESTUs and PTs shows that some

ESTUs are preferentially associated with only one PT, whileothers present a much larger pigment diversity. ESTU IIA, themost abundant and ubiquitous ESTU in the field (5, 6), displaysthe widest PT diversity (Fig. 4B), a finding confirmed by clade IIisolates spanning the largest diversity of pigment content, withrepresentative strains of PT 2A, 3a, 3c, and 3dB within this clade(Dataset S3) (7, 11, 50–52). This suggests that this ESTU cancolonize all light color niches, an ability that might be partiallyresponsible for its global ecological success. Our current resultsdo not support the previously observed correlation betweenclade III and PT 3a (29) since the two ESTUs defined within thisclade (IIIA and B) were associated with PT 3c and/or 3f. Thisdiscrepancy could be due either to the different methods used inthese studies or to the occurrence of genetically distinct clade IIIpopulations in coastal areas of the northwestern Pacific Oceanand along the Tara Oceans transect. However, the pigmentphenotype of strains isolated to date is more consistent with ourfindings (Dataset S3) (16, 36).In contrast to most other PTs, the association between PT 3dA

and ESTUs was found to be nearly exclusive in the field, asESTUs from clades I, IV, CRD1, and EnvA were not associatedwith any other PT, and reciprocally PT 3dA is only associatedwith these clades (Fig. 4A). An interesting exception to thisgeneral rule was observed in the Benguela upwelling (TARA_067), where the dominant ESTU IA population both displays a3a mpeBA allele and possesses fciA/B and mpeZ genes (Figs. S6Kand S7), suggesting that cells, which were initially green-lightspecialists (PT 3a), have inherited a complete CA4-A islandthrough lateral gene transfer at this station. Interestingly, amongthe seven clade I strains sequenced to date, three possess a 3ampeBA allele, among which WH8016 also has a CA4-A islandbut only partial (lacking mpeZ) and therefore not functional(11). It is thus difficult to conclude whether the lateral transfer ofthis island, likely a rare event since it was only observed inpopulations of the Benguela upwelling, has conferred thesepopulations the ability to perform CA4.

Grébert et al. PNAS Latest Articles | 7 of 10

ENVIRONMEN

TAL

SCIENCE

SPN

ASPL

US

Another important result of this study was the unsuspectedimportance of populations that have likely lost the ability tochromatically acclimate, specifically in warm HNLC areas, whichcover wide expanses of the South Pacific Ocean (53). In-terestingly, populations living in these ultra-oligotrophic envi-ronments have a different genetic basis for their consistentlyelevated PUB phenotype than do typical blue-light specialists(i.e., PT 3c), since they have lost the CA4 regulators fciA/B andaccumulated mutations in mpeY, a yet-uncharacterized memberof the phycobilin lyase family, as observed in strain BIOS-E4-1(Fig. S6K) (11). This finding, consistent with the previous ob-servation that the south Pacific is dominated by high-PUBSynechococcus (22), is further supported by the recent se-quencing of three isolates from the Equatorial Pacific, strainsMITS9504, MITS9509 (both CRD1C), and MITS9508 (CRD1A)(54), all of which contain, like BIOS-E4-1, a 3dA mpeBA allele, aCA4-A island lacking fciA/B, and a partial (MITS9508) or highlydegenerated (two other MIT strains) mpeY gene sequence (Fig.S6K). Thus, these natural CA4-A mutants seem to have adaptedto blue, ultra-oligotrophic waters by inactivating a likely ener-getically costly acclimation mechanism (positive selection), al-though we cannot exclude that it might be a consequence of thelower selection efficiency associated to the reduced effectivepopulation size of Synechococcus in such an extreme environ-ment (genetic drift). If, as we hypothesize, all Synechococcus cellscounted as 3dA at these stations are CA4-deficient, these naturalmutants would represent about 15% of the total 3dA population.In contrast, CRD1-A populations of the eastern border of thePacific Ocean (TARA_102, 109–110, 137) are likely true CA4populations as they possess all CA4 genes (Fig. S6K).In conclusion, our study provided insights into the distribution,

ecology, and adaptive value of all known Synechococcus PTs.Unexpectedly, the sum of 3dA and 3dB constituted about 40% ofthe total Synechococcus counts in the Tara Oceans dataset,making chromatic acclimaters (PT 3d) the most globally abun-dant PT, even when taking into account potential CA4-deficientnatural mutants. In addition, this PT made up 95% of theSynechococcus population at high latitudes and was present inevery one of the five major clades in the field (I, II, III, IV, andCRD1). This suggests that chromatic acclimation likely confers astrong adaptive advantage compared with strains with a fixedpigmentation, particularly in vertically mixed environments andat depth at stations with a stratified water column. The occur-rence of natural CA4 mutants and evidence for lateral transfer ofthe CA4 genomic island further support previous hypothesesthat not only temperature and nutrient availability (3, 5, 6) butalso light quality (7, 52) coexert selective pressures affectingmarine Synechococcus evolution. Thus, changes in pigment di-versity could occur in response to changes in light niches by ac-quisition or loss of specific PBS synthesis and/or regulationgenes, as previously observed for phosphorus and nitrogentransport genes in Prochlorococcus (55–57). Still, the complexinteractions between PTs, vertical phylogeny, and environmentalparameters remain unclear and more work is needed to refineour understanding of the balance between the forces shapingcommunity composition and Synechococcus evolution. At theboundaries of Synechococcus environmental niche(s), where theharshest conditions are encountered, both pigment and cladediversity are drastically reduced, and this concomitant reductiontends to support a coselection by light quality and other envi-ronmental parameters. On the contrary, the diverse PTs occur-ring within some clades, as well as the co-occurrence of differentPTs at most stations compared with more clearcut clade shifts(e.g., in the Red Sea/Indian Ocean) might indicate that lightquality is not the strongest selective force or that light changesare too transient to allow the dominance and fixation of a par-ticular PT in a population. Future experimental work exploringthe fitness of distinct ESTU/PT combinations under different

controlled environmental conditions (including temperature,nutrients, and light) might help clarifying the respective effects ofthese parameters on the diversification of this ecologically im-portant photosynthetic organism.

Materials and MethodsMetagenomic Samples. This study focused on 109 metagenomic samplescorresponding to 65 stations from the worldwide oceans collected during the2.5-y Tara Oceans circumnavigation (2009–2011). Water sample and se-quence processing are the same as in ref. 6. Dataset S4 describes all meta-genomic samples with location and sequencing effort. Sequencing depthsranged from 16 × 106 to 258 × 106 reads per sample after quality control andpaired-reads merging, and corresponding fragments lengths averaged 164 ±20 bp (median, 168 bp).

Databases: Reference and Outgroup Sequences. A reference database com-prising the full-length gene or operon nucleotide sequences was generatedfor each marker used in this study (cpcBA, mpeBA, and mpeW) based onculture isolates with characterized pigment type (Dataset S1). These data-bases comprised 83 cpcBA sequences (64 unique), including 18 PT 1, 5 PT 2A,19 PT 2B, and 39 PT 3; 41mpeBA sequences (all unique), including 11 PT 3a, 2PT 3f, 11 PT 3dA, and 17 PT 3dB; and 5 unique mpeW sequences. For eachmarker, a reference alignment was generated with MAFFT L-INS-i, version6.953b (58), and a reference phylogenetic tree was inferred with PhyML,version 20120412 (GTR + I + G, 10 random starting trees, best of SPR and NNImoves, 500 bootstraps) (59); and drawn using the ETE Toolkit (60).

A database of outgroups was also built, comprising paralogous sequencesfrom marine Synechococcus or Prochlorococcus as well as orthologous se-quences from other marine and freshwater organisms retrieved from publicdatabases. For cpcBA and mpeBA, the outgroup databases comprised apcA,apcB, apcD, apcF, and cpeBA from marine Synechococcus, ppeBA fromProchlorococcus, cpcBA and cpeBA from other nonpicocyanobacterial or-ganisms as well as either mpeBA or cpcBA from marine Synechococcus, re-spectively (Datasets S1 and S2). FormpeW, the outgroup database was madeof paralogous genes (mpeZ,mpeY, and cpeY) from marine Synechococcus orProchlorococcus, as no ortholog could be identified in public databases.Similarly, for mpeY and mpeZ, the outgroup database comprised cpeY,mpeW, as well as mpeZ or mpeY, respectively. The outgroup database formpeU comprised cpeF paralogous sequences from marine Synechococcusand Prochlorococcus. No outgroup database was used for fciAB, as noparalogs or other distantly related sequences were found either in marineSynechococcus and Prochlorococcus or in public databases.

Read Assignation and Estimation of PT Abundance. Reads were preselectedusing BLAST+ (61) with relaxed parameters (blastn, maximum E-value of1e-5, minimum percent identity of 60%, minimum 75% of read lengthaligned), using reference sequences as subjects; the selection was then re-fined by a second BLAST+ round against databases of outgroups: reads witha best hit to outgroup sequences were excluded from downstream analysis.Selected reads were then aligned to the marker reference alignment withMAFFT, version 7.299b (–addfragments –adjustdirectionaccurately), andplaced in the marker reference phylogenetic tree with pplacer (62). For eachread, pplacer returns a list of possible positions (referred to as placements) atwhich it can be placed in the tree and their associated “likelihood weightratio” (LWR) (proxy for the probability of the placement; see pplacer pub-lication and documentation for more details). Reads were then assigned to apigment type using a custom classifier written in Python. Briefly, internalnodes of the reference tree were assigned a pigment type based on thepigmentation of descending nodes (PT of child reference sequences if thesame for all of them, “unclassified” otherwise). For each read, placementswere assigned to their nearest ascending or descending node based on theirrelative position on the edge, and the lowest common ancestor (LCA) of theset of nodes for which the cumulated LWR was greater than 0.95 (LCA ofpossible placements at 95% probability) was then computed. Finally, theread was assigned to the pigment type of this LCA. Different combinationsof read assignment parameters (LCA at 90%, 95%, or 100%; assignation ofplacements to the ascending, descending, or nearest node) were alsoassessed, and resulted either in higher rates of unassigned reads or ofwrongly assigned reads (Fig. S2).

Read counts were normalized by adjusted marker length: for each markerand each sequence file, counts were normalized by (L − ℓ + 1), with L, thelength of the marker gene (cpcBA mean length: 1,053.7 bp; mpeBA meanlength: 1,054.6 bp; mpeW mean length: 1,193.3 bp), and ℓ, the mean lengthof reads in the sequence file. Finally, the abundance of PT 1, 2A, and 2B was

8 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1717069115 Grébert et al.

defined as the normalized cpcBA read counts of these PTs, the abundance ofPT 3a, 3f, and 3dA as the normalizedmpeBA read counts of these PTs, 3dB asthe normalized mpeW count and 3c as the difference between the nor-malized mpeBA (3c + 3dB) read count and the PT 3dB count assessed withmpeW. The abundance of unclassified sequences was also taken into ac-count. Detailed petB counts for clade and ESTU abundances were obtainedfrom ref. 6.

Read Assignment Simulations. For each marker, simulated reads were gen-erated from one reference sequence at a time using a sliding window of 100,125, or 150 bp (Tara Oceans mean read length, 164.2 bp; median, 169 bp)and steps of 5 bp. Simulated reads were then assigned to a pigment typewith the aforementioned bioinformatic pipeline, using all reference se-quences except the one used to simulate reads (“leave-one-out” cross-validation scheme). Inferred PTs of simulated fragments were then com-pared with known PTs of reference sequences.

Statistical Analyses.All environmental parameters used for statistical analysesare the same as in ref. 6, except the blue to green irradiance ratio that wasmodeled as described in SI Materials and Methods. Hierarchical clusteringand NMDS analyses of stations were performed using R (63) packages clus-ter, version 1.14.4 (64) and MASS, version 7.3-29 (65), respectively. PT con-tingency tables were filtered by considering only stations with more than 30cpcBA reads and 30 mpeBA reads, and only PT appearing in at least twostations and with more than 150 reads in the whole dataset. Contingencytables were normalized using Hellinger transformation that gives lowerweights to rare PT. The Bray–Curtis distance was then used for ordination(isoMDS function; maxit, 100; k, 2). Correlations were performed with R

package Hmisc_3.17–4 with Benjamini and Hochberg multiple-comparisonadjusted P value (66).

ACKNOWLEDGMENTS. We warmly thank Dr. Annick Bricaud for fruitful dis-cussions on biooptics, members of the Analysis and Bioinformatics for Ma-rine Science (ABiMS) platform (Roscoff) for providing us an efficient storageand computing facility for our bioinformatics analyses, as well as the NaturalEnvironment Research Council Biomolecular Analysis Facility (Centre for Ge-nomic Research, University of Liverpool) for sequencing some Synechococcusgenomes used in this study. We also thank the support and commitment ofthe Tara Oceans coordinators and consortium, Agnès b. and E. Bourgois, theVeolia Environment Foundation, Région Bretagne, Lorient Agglomeration,World Courier, Illumina, the Electricité de France (EDF) Foundation, Fonda-tion pour la Recherche sur la Biodiversité, the Prince Albert II de MonacoFoundation, and the Tara schooner and its captains and crew. Tara Oceanswould not exist without continuous support from 23 institutes (https://oceans.taraexpeditions.org/en/). This work was supported by the French“Agence Nationale de la Recherche” Programs SAMOSA (Synechococcus asa model genus for studying adaptation of marine phytoplankton to envi-ronmental changes) (Grant ANR-13-ADAP-0010) and France Génomique(Grant ANR-10-INBS-09), the French Government “Investissements d’Avenir”programs World Ocean Bioressources, Biotechnologies and Earth-SystemServices (OCEANOMICS) (Grant ANR-11-BTBR-0008), the European Union’sSeventh Framework Programs FP7 MicroB3 (Grant Agreement 287589),and Marine Microorganisms: Cultivation Methods for Improving Their Bio-technological Applications (Macumba; Grant Agreement 311975), UK Natu-ral Environment Research Council Grant NE/I00985X/1, and the SpanishMinistry of Science and Innovation Grant MicroOcean PANGENOMICS(GL2011-26848/BOS). This article is contribution number 69 of Tara Oceans.

1. Guidi L, et al.; Tara Oceans Coordinators (2016) Plankton networks driving carbonexport in the oligotrophic ocean. Nature 532:465–470.

2. Flombaum P, et al. (2013) Present and future global distributions of the marinecyanobacteria Prochlorococcus and Synechococcus. Proc Natl Acad Sci USA 110:9824–9829.

3. Zwirglmaier K, et al. (2008) Global phylogeography of marine Synechococcus andProchlorococcus reveals a distinct partitioning of lineages among oceanic biomes.Environ Microbiol 10:147–161.

4. Mazard S, Ostrowski M, Partensky F, Scanlan DJ (2012) Multi-locus sequence analysis,taxonomic resolution and biogeography of marine Synechococcus. Environ Microbiol14:372–386.

5. Sohm JA, et al. (2016) Co-occurring Synechococcus ecotypes occupy four major oce-anic regimes defined by temperature, macronutrients and iron. ISME J 10:333–345.

6. Farrant GK, et al. (2016) Delineating ecologically significant taxonomic unitsfrom global patterns of marine picocyanobacteria. Proc Natl Acad Sci USA 113:E3365–E3374.

7. Six C, et al. (2007) Diversity and evolution of phycobilisomes in marine Synechococcusspp.: A comparative genomics study. Genome Biol 8:R259.

8. Alberte RS, Wood AM, Kursar TA, Guillard RRL (1984) Novel phycoerythrins in marineSynechococcus spp.: Characterization and evolutionary and ecological implications.Plant Physiol 75:732–739.

9. Ong LJ, Glazer AN (1991) Phycoerythrins of marine unicellular cyanobacteria. I. Bilintypes and locations and energy transfer pathways in Synechococcus spp. phycoery-thrins. J Biol Chem 266:9515–9527.

10. Sidler WA (1994) Phycobilisome and phycobiliprotein structures. The Molecular Bi-ology of Cyanobacteria, Advances in Photosynthesis (Springer, Dordrecht, TheNetherlands), pp 139–216.

11. Humily F, et al. (2013) A gene island with two possible configurations is involved inchromatic acclimation in marine Synechococcus. PLoS One 8:e84459.

12. Palenik B (2001) Chromatic adaptation in marine Synechococcus strains. Appl EnvironMicrobiol 67:991–994.

13. Everroad C, et al. (2006) Biochemical bases of type IV chromatic adaptation in marineSynechococcus spp. J Bacteriol 188:3345–3356.

14. Shukla A, et al. (2012) Phycoerythrin-specific bilin lyase-isomerase controls blue-greenchromatic acclimation in marine Synechococcus. Proc Natl Acad Sci USA 109:20136–20141.

15. Sanfilippo JE, et al. (2016) Self-regulating genomic island encoding tandem regulatorsconfers chromatic acclimation to marine Synechococcus. Proc Natl Acad Sci USA 113:6077–6082.

16. Toledo G, Palenik B, Brahamsha B (1999) Swimming marine Synechococcus strainswith widely different photosynthetic pigment ratios form a monophyletic group.Appl Environ Microbiol 65:5247–5251.

17. Humily F, et al. (2014) Development of a targeted metagenomic approach to study agenomic region involved in light harvesting in marine Synechococcus. FEMS MicrobiolEcol 88:231–249.

18. Jiang T, et al. (2016) Temporal and spatial variations of abundance of phycocyanin-and phycoerythrin-rich Synechococcus in Pearl River Estuary and adjacent coastalarea. J Ocean Univ China 15:897–904.

19. Olson RJ, Chisholm SW, Zettler ER, Armbrust EV (1990) Pigments, size, and distribu-tions of Synechococcus in the North Atlantic and Pacific Oceans. Limnol Oceanogr 35:45–58.

20. Sherry ND, Wood AM (2001) Phycoerythrin-containing picocyanobacteria in theArabian Sea in February 1995: Diel patterns, spatial variability, and growth rates.Deep Sea Res Part II Top Stud Oceanogr 48:1263–1283.

21. Lantoine F, Neveux J (1997) Spatial and seasonal variations in abundance and spectralcharacteristics of phycoerythrins in the tropical northeastern Atlantic Ocean. DeepSea Res Part I Oceanogr Res Pap 44:223–246.

22. Neveux J, Lantoine F, Vaulot D, Marie D, Blanchot J (1999) Phycoerythrins in thesouthern tropical and equatorial Pacific Ocean: Evidence for new cyanobacterialtypes. J Geophys Res Oceans 104:3311–3321.

23. Campbell L, et al. (1998) Response of microbial community structure to environmentalforcing in the Arabian Sea. Deep Sea Res Part II Top Stud Oceanogr 45:2301–2325.

24. Wood AM, Lipsen M, Coble P (1999) Fluorescence-based characterization ofphycoerythrin-containing cyanobacterial communities in the Arabian Sea during theNortheast and early Southwest Monsoon (1994–1995). Deep Sea Res Part II Top StudOceanogr 46:1769–1790.

25. Yona D, Park MO, Oh SJ, Shin WC (2014) Distribution of Synechococcus and its phy-coerythrin pigment in relation to environmental factors in the East Sea, Korea. OceanSci J 49:367–382.

26. Hoge FE, Wright CW, Kana TM, Swift RN, Yungel JK (1998) Spatial variability ofoceanic phycoerythrin spectral types derived from airborne laser-induced fluores-cence emissions. Appl Opt 37:4744–4749.

27. Wood AM, Phinney DA, Yentsch CS (1998) Water column transparency and the dis-tribution of spectrally distinct forms of phycoerythrin-containing organisms. Mar EcolProg Ser 162:25–31.

28. Campbell L, Iturriaga R (1988) Identification of Synechococcus spp. in the Sargasso Seaby immunofluorescence and fluorescence excitation spectroscopy performed on in-dividual cells. Limnol Oceanogr 33:1196–1201.

29. Xia X, et al. (2017) Phylogeography and pigment type diversity of Synechococcuscyanobacteria in surface waters of the northwestern pacific ocean. Environ Microbiol19:142–158.

30. Xia X, Liu H, Choi D, Noh JH (2017) Variation of Synechococcus pigment genetic di-versity along two turbidity gradients in the China Seas. Microb Ecol 75:10–21.

31. Xia X, Guo W, Tan S, Liu H (2017) Synechococcus assemblages across the salinitygradient in a salt wedge estuary. Front Microbiol 8:1254.

32. Liu H, Jing H, Wong THC, Chen B (2014) Co-occurrence of phycocyanin- andphycoerythrin-rich Synechococcus in subtropical estuarine and coastal waters of HongKong. Environ Microbiol Rep 6:90–99.

33. Chung C-C, Gong G-C, Huang C-Y, Lin J-Y, Lin Y-C (2015) Changes in the Synechococcusassemblage composition at the surface of the East China Sea due to flooding of theChangjiang river. Microb Ecol 70:677–688.

34. Haverkamp T, et al. (2008) Diversity and phylogeny of Baltic Sea picocyanobacteriainferred from their ITS and phycobiliprotein operons. Environ Microbiol 10:174–188.

35. Stomp M, et al. (2007) Colourful coexistence of red and green picocyanobacteria inlakes and seas. Ecol Lett 10:290–298.

36. Hunter-Cevera KR, Post AF, Peacock EE, Sosik HM (2016) Diversity of Synechococcus atthe Martha’s Vineyard coastal observatory: Insights from culture isolations, clone li-braries, and flow cytometry. Microb Ecol 71:276–289.

37. Fuller NJ, et al. (2003) Clade-specific 16S ribosomal DNA oligonucleotides reveal thepredominance of a single marine Synechococcus clade throughout a stratified watercolumn in the Red Sea. Appl Environ Microbiol 69:2430–2443.

Grébert et al. PNAS Latest Articles | 9 of 10

ENVIRONMEN

TAL

SCIENCE

SPN

ASPL

US

38. Larsson J, et al. (2014) Picocyanobacteria containing a novel pigment gene clusterdominate the brackish water Baltic Sea. ISME J 8:1892–1903.

39. Chen F, et al. (2004) Phylogenetic diversity of Synechococcus in the Chesapeake Bayrevealed by ribulose-1,5-bisphosphate carboxylase-oxygenase (RuBisCO) large subunitgene (rbcL) sequences. Aquat Microb Ecol 36:153–164.

40. Choi DH, Noh JH (2009) Phylogenetic diversity of Synechococcus strains isolated fromthe East China Sea and the East Sea. FEMS Microbiol Ecol 69:439–448.

41. Sunagawa S, et al.; Tara Oceans Coordinators (2015) Ocean plankton. Structure andfunction of the global ocean microbiome. Science 348:1261359.

42. Logares R, et al. (2014) Metagenomic 16S rDNA Illumina tags are a powerful alter-native to amplicon sequencing to explore diversity and structure of microbial com-munities. Environ Microbiol 16:2659–2671.

43. Pearman PB, Guisan A, Broennimann O, Randin CF (2008) Niche dynamics in space andtime. Trends Ecol Evol 23:149–158.

44. Paulsen ML, et al. (2016) Synechococcus in the Atlantic gateway to the Arctic Ocean.Front Mar Sci 3:191.

45. Haverkamp THA, et al. (2009) Colorful microdiversity of Synechococcus strains(picocyanobacteria) isolated from the Baltic Sea. ISME J 3:397–408.

46. Cabello-Yeves PJ, et al. (2017) Novel Synechococcus genomes reconstructed fromfreshwater reservoirs. Front Microbiol 8:1151.

47. Mahmoud RM, et al. (2017) Adaptation to blue light in marine Synechococcus re-quires MpeU, an enzyme with similarity to phycoerythrobilin lyase isomerases. FrontMicrobiol 8:243.

48. Veldhuis MJW, Kraay GW (1993) Cell abundance and fluorescence of picoplankton inrelation to growth irradiance and nitrogen availability in the Red Sea. Neth J Sea Res31:135–145.

49. Katano T, Nakano S (2006) Growth rates of Synechococcus types with differentphycoerythrin composition estimated by dual-laser flow cytometry in relationship tothe light environment in the Uwa Sea. J Sea Res 55:182–190.

50. Ahlgren NA, Rocap G (2006) Culture isolation and culture-independent clone librariesreveal new marine Synechococcus ecotypes with distinctive light and N physiologies.Appl Environ Microbiol 72:7193–7204.

51. Bemal S, Anil AC (2016) Genetic and ecophysiological traits of Synechococcus strains isolatedfrom coastal and open ocean waters of the Arabian Sea. FEMS Microbiol Ecol 92:fiw162.

52. Everroad RC, Wood AM (2012) Phycoerythrin evolution and diversification of spectralphenotype in marine Synechococcus and related picocyanobacteria. Mol PhylogenetEvol 64:381–392.

53. Morel A, et al. (2007) Optical properties of the “clearest” natural waters. Limnol

Oceanogr 52:217–229.54. Cubillos-Ruiz A, Berta-Thompson JW, Becker JW, van der Donk WA, Chisholm SW

(2017) Evolutionary radiation of lanthipeptides in marine cyanobacteria. Proc Natl

Acad Sci USA 114:E5424–E5433.55. Martiny AC, Huang Y, Li W (2009) Occurrence of phosphate acquisition genes

in Prochlorococcus cells from different ocean regions. Environ Microbiol 11:

1340–1347.56. Martiny AC, Coleman ML, Chisholm SW (2006) Phosphate acquisition genes in

Prochlorococcus ecotypes: Evidence for genome-wide adaptation. Proc Natl Acad Sci

USA 103:12552–12557.57. Martiny AC, Kathuria S, Berube PM (2009) Widespread metabolic potential for nitrite

and nitrate assimilation among Prochlorococcus ecotypes. Proc Natl Acad Sci USA 106:

10787–10792.58. Katoh K, Standley DM (2013) MAFFT multiple sequence alignment software version 7:

Improvements in performance and usability. Mol Biol Evol 30:772–780.59. Guindon S, et al. (2010) New algorithms and methods to estimate maximum-

likelihood phylogenies: Assessing the performance of PhyML 3.0. Syst Biol 59:

307–321.60. Huerta-Cepas J, Serra F, Bork P (2016) ETE 3: Reconstruction, analysis, and visualiza-

tion of phylogenomic data. Mol Biol Evol 33:1635–1638.61. Camacho C, et al. (2009) BLAST+: Architecture and applications. BMC Bioinformatics

10:421.62. Matsen FA, Kodner RB, Armbrust EV (2010) pplacer: Linear time maximum-likelihood

and Bayesian phylogenetic placement of sequences onto a fixed reference tree. BMC

Bioinformatics 11:538.63. R Core Team (2014) R: A language and environment for statistical computing (R

Found Stat Comput, Vienna). Available at www.R-project.org/.64. Maechler M, Rousseeuw P, Struyf A, Hubert M, Hornik K (2017) cluster: Cluster

analysis basics and extensions. R package, version 2.0.6. Available at: http://CRAN.R-

project.org/package=cluster. Accessed April 18, 2016.65. Venables WN, Ripley BD (2002)Modern Applied Statistics with S (Springer, New York),

4th Ed.66. Harrell FE (2016) Hmisc: Harrell Miscellaneous. Available at CRAN.R-project.org/

package=Hmisc. Accessed April 18, 2016.

10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1717069115 Grébert et al.